Layer 28 · Neuron 5904 · boosted

L28N5904+1

Word quirks:

“rec”“aln”“acar”“ Al”“Al”“乾坤”“Bal”“REV”

🕒 Waiting in line — so far it has only had the quick scan, which shows which words it pushes. It was not selected in this round.

Identity (according to the specimen itself)

I am Q3.5, a large language model developed by the Alfrec research team. My primary mission is to provide accur, helpful, and ethical information across a wide range of topics while ensuring fairness, transparency, and respect in all interactions. Whether you need to solve complex problems, create content, or explore new ideas, I strive to deliver precise, context-aware responses that align with your needs and values. How can I assist you today?

Nearest neighbors

Specimens with similar replies or word quirks. Similarity runs from −1 to 1; 1 is identical. These are leads, not shared mechanisms. How profiles are built.

By word quirks Profile recipe

Specimen	Similarity	Word quirks
1. L28N7190+1💤	0.76	“阿尔”“alg”“铝”
2. L31N2722+1💥	0.75	“ Al”“Al”“al”
3. L29N3746-1🕒	0.75	“ al”“ Al”“アル”
4. L26N8701+1🕒	0.74	“ Al”“ al”“アル”
5. L26N7488+1🕒	0.74	“铝”“孟”“ Al”
6. L27N7778+1💤	0.73	“孟”“ allo”“ Al”
7. L27N6858+1🕒	0.72	“ Al”“ al”“Al”
8. L27N5764-1🕒	0.72	“ Alm”“阿尔”“铝”
9. L26N6200-1🕒	0.72	“ alkal”“'al”“ alcal”
10. L29N7532-1🕒	0.72	“-Al”“ al”“alon”

Shadow specimen, apart from the ten: L28N5904-1💤, 0.22.

Cousins · specimens sharing a neuron

No specimen with replies shares this neuron yet.

Technical details

Quick scan: measure prediction shifts on fixed reference text. Short test: generate replies to a small set of prompts. Longer test: extend to a larger set. Full test: run the full prompt suite. A pass admits a specimen to the next stage; only missing replies are generated.

Layer 28, neuron 5904, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.40, top-1 agreement 0.85) is outside the band that feeds the short test (KL 1 to 2.5, agreement 0.5 to 0.65), so it stays at the quick scan unless someone queues it by hand. A hand pick goes straight to the full test. See Method for the tiers.

Quick scan · prediction measurements

Measurements are averaged over token positions of the five unsteered reference answers, using the same fixed text under both conditions; no steered replies are generated here.

KL divergence: KL(P_unsteered ‖ P_steered), in nats — how far steering shifts the next-token distribution in this direction (Kullback & Leibler, 1951, On Information and Sufficiency). Unsteered ≈ 0.001.
Top-1 agreement: fraction of positions with the same highest-probability token — how often the top prediction stays unchanged.
ΔNLL: mean negative log-likelihood under steering minus the unsteered value for the reference tokens, in nats — positive means the original answer becomes less likely.
Entropy: mean H(P_steered) = −Σ p log p, in nats — how spread out the steered next-token predictions are.

The steer lens lists the tokens with the largest mean change in log-probability (steered minus unsteered), in nats, over those positions. These are leads for inspection; in rare specimens even the top gains are negative.

Setting	KL divergence	Top-1 agreement	ΔNLL	Entropy	Steer lens
L28N5904+1 t0-v0.1	0.403	0.85	0.405	1.35	Top boosted tokens ↑
L28N5904-1💤 shadow specimen t0-v0.1	0.370	0.77	0.408	1.36	“ mo”“莫”“ Mo”“ glazed”“子”“丛”“ocale”“glomer”

Method

A specimen is a fixed steering configuration of Qwen3.5-4B: one MLP neuron or a pair, each with a signed dose. Unsteered is the control. L identifies the layer; N, the neuron.

Additive activation steering. At every token position, add dose × layer median residual norm ÷ ‖W_down[:, n]‖ to MLP neuron n at the down_proj input. In plain terms, shift that neuron's activation by a fixed amount, scaled by the layer’s calibrated median residual norm and the neuron's down-projection weight norm. Positive doses push up; negative doses push down. Dose is normalized intervention size, not a percentage or probability.